Model projections of future climate change require that coupled climate-vegetation models are developed and validated, i.e. these models should be able to reproduce past climate and vegetation change. Records of pollen deposited in lake bottoms and peat bogs can provide the information needed to validate these models. The aim of this thesis was i) to explore the modern relationships between pollen and vegetation in northern and temperate China and estimate pollen productivity of major plant taxa, and ii) to use the results of i) to produce the first reconstruction of plant cover in China over the last 10 000 years for the purpose of climate modelling. A study of the modern pollen-vegetation-climate relationships was performed in northwestern China (Paper I). Pollen productivity for 18 major plants of cultural landscapes in central-eastern China was estimated (Paper II). Based on a synthesis and evaluation of all existing estimates of pollen productivity in the study region, a standard dataset of pollen productivity for 31 plant taxa is proposed (Paper III). This dataset was used to achieve pollen-based REVEALS reconstructions of plant cover over the last 10 000 years in 35 regions of northern and temperate China (Paper IV). The major findings can be summarized as follows. Paper I: Annual precipitation (Pann) is the major climatic factor influencing pollen assemblages, followed by July precipitation (PJul). The shared effect of combinations of two climatic factors explains a larger portion of the variation in pollen data than individual variables. Paper II: Of the 16 reliable pollen productivities estimated, the estimates for 8 taxa are new, Castanea, Cupressaceae, Robinia/Sophora, Anthemis type/Aster type, Cannabis/Humulus, Caryophyllaceae, Cruciferae, and Galium type. Trees have in general larger pollen productivity than herbs. Paper III: Of the total 31 taxa for which estimates of pollen productivity are available in China, 13 taxa have more than 1 value. All or most of these values are similar for Artemisia, Cyperaceae, Larix, Quercus and Pinus. Eight taxa have very variable estimates. Paper IV: The REVEALS plant percentage-cover strongly differs from the pollen percentages, and they provide new important insights on past changes in plant composition and vegetation dynamics.

The Ethiopian Wildlife Conservation Organization has little information on the lll extent and ecological determinants of plant species composition and distribution in Omo National Park. Elsewhere, the determinants of African savanna vegetation dynamics have been the focus of numerous investigations, yet our understanding of the hierarchical nature and relative importance of any relationships remains very general. As well, our ability to derive predictions about vegetation responses is limited to extreme generalizations. African savanna landscape ecotones have received even less attention than most landscapes. In this dissertation, I test hypotheses about plant species distribution-determinant relationships in Omo National Park, a park that occurs in a landscape ecotone. Determinants investigated include rainfall and topographic-related gradients in soil moisture and edaphic conditions, and availability of soil constituents. Rainfall pattern was determined from meteorological data and multiple linear regression. Topographic attributes were measured in the field. Availability of soil constituents was determined by evaluating and using spatial interpolation models using limited soils data, and construction of surface soil maps. Hypotheses were tested using the simple and partial Mantel tests of matrix association. Results demonstrate that predictions using spatial interpolation models based on limited, coarse-scale soils data are accurate and reliable when compared with more data-intensive investigations. Results using spatial statistics indicate that the nature of the spatial pattern of perennial species associations is a monotonic spatial trend. The distribution of perennial species associations is influenced both directly or indirectly by rainfall. An indirect rainfall relationship occurs when there is variability in topography. The means of influence by the topographic-related attributes is unclear despite significant Mantel results. In the topographically invariant portion of the study area, however, exchangeable sodium and magnesium concentrations appear to indirectly influence the distribution of perennial species associations. Strength and ordering of the vegetation-determinant relationships varied depending on the type of perennial species association type being investigated. Differences in species range of tolerance and rate of change in species association, depending on association type, landforrn, and parent material differences, account for the nature of the vegetation-determinant relationships, the ordering of the determinants, and variability in responses.

Bird-vegetation associations are a base for bird conservation and management, as well as for predictions of the effects of resource management and climate change on wildlife populations. A recent shift in forest management priorities from timber production to native species' habitat conservation on federal lands has emphasized the need to understand the mechanisms underlying the effects of vegetation management on wildlife. The assumption of strong bird-vegetation relationships is rarely tested for forest birds, especially at large temporal extents, which are more likely to reveal instabilities in bird-vegetation relationships than short-term studies. This study aimed to quantify bird-vegetation relationships and investigate their strength in young thinned and unthinned Douglas-fir forest stands over ten to years post thinning. Additionally, this study investigated whether disturbance associated with forest thinning decoupled bird-vegetation relationships in the thinned and unthinned stands. I used abundance or occurrence data for eight bird species collected at 58 point count surveys, conducted during six breeding seasons over ten years following forest thinning. I obtained detailed local-scale vegetation characteristics associated with bird sampling points and modeled bird occurrence or abundance as a function of vegetation characteristics. Vegetation characteristics explaining individual species occurrence or abundance varied among species and among years for any given species. Six out of eight species showed responses to examined vegetation characteristics. For three out of six species, the effects of vegetation characteristics on bird occurrence or abundance remained consistently positive or negative over time. For the other three species the absolute effect of vegetation decreased over time to that of not statistically different from zero. The estimates of vegetation effects on bird response varied in size among years, though they were not statistically different among years. Magnitude of vegetation effect on bird occurrence or abundance did not increase with time, nor was it related to species prevalence or abundance. I found evidence of a response threshold for one species, Swainson's thrush. I suggest that changing abundance of resources, associated with thinning disturbance, demographic stochasticity associated with small population sizes, as well as large-scale demographic processes and possibly life history traits of examined species, mediate the strength of local-level bird-vegetation associations. Variability of vegetation effects on bird occurrence or abundance over time suggests a greater uncertainty of results of vegetation-related wildlife management efforts than has previously been suggested by short-term studies. Therefore, incorporation of the uncertainty of bird-vegetation relationships into predictive models and continued long-term monitoring of species response to management may be crucial for future successful and effective management decisions. / Graduation date: 2012

Comprendre la complexité des relations entre nature et sociétés a toujours constitué un défi pour les recherches dans le domaine de l'environnement. A l'ère de la compilation de données dans de nombreuses disciplines, la modélisation constitue l'une des options les plus attractives pour comprendre et traduire les évolutions du milieu naturel. Les objectifs principaux de ce travail de recherche sont 1) de restituer la structure du couvert végétal régional et d'en retracer les évolutions au cours de l'Holocène au sein de deux régions du nord-ouest de la France, le Bassin parisien et le Massif armoricain, et 2) de les interpréter en termes de relations entre les sociétés pré- et protohistoriques et leur milieu naturel. Pour cela, nous avons mis en place une démarche méthodologique visant à 1) redéfinir le cadre chronologique des séquences polliniques rassemblées pour les deux régions, 2) tester, adapter puis appliquer un modèle de reconstitution quantitative du couvert végétal, 3) définir une histoire de la végétation estimée relative à chaque aire géographique modélisée, 4) comparer ces histoires régionales et 5) les confronter aux informations disponibles en matière de densités de sites archéologiques et de proximité de ces sites aux points de sondages polliniques. Des cadres structurant l'évolution du couvert végétal pour chacune des deux régions étudiées ressortent les grandes étapes de la mise en place des paysages au cours de l'Holocène. Des spécificités régionales ont cependant pu être identifiées, tant en matière de chronologie de l'évolution de la végétation estimée qu'en matière de proportions relatives des taxons présents. Une certaine singularité doit donc être reconnue à l'évolution des paysages holocènes de chacune de ces deux parties du nord-ouest de la France. On remarque notamment des dynamiques et des niveaux d'anthropisation tout à fait distincts entre Bassin parisien et Massif armoricain de la fin du Mésolithique à celle du Néolithique. A l'inverse, on retrouve de part et d'autre une certaine similarité entre les proportions estimées des taxons marqueurs d'anthropisation entre la fin du Néolithique et celle de l'âge du Bronze. Des liens évidents entre les variations de l'impact anthropique enregistré et celles de la densité des occupations reconnues ont par ailleurs pu être observés. / Understanding the complex relationship between nature and societies has always been a challenge for environmental sciences. With the growing impact of databases in many disciplines, the modelling appears as one of the most attractive methods for understanding and describing environmental changes. The main objectives of this research work are 1) to reconstruct the structure of the regional vegetation cover and to determine its trends during the Holocene in two regions of north-western France, the Paris Basin and the Armorican Massif, and 2) to interpret them in terms of relationships between the pre- and proto-historic societies and their environment. Thus, we set up a methodological approach aimed at 1) redefining the chronological framework of pollen sequences collected for both regions, 2) testing, adapting and then applying a model in order to produce quantitative estimates of vegetation cover, 3) defining a history of the estimated vegetation for each area modelled, 4) comparing these regional results and 5) confronting them to the available informations in terms of densities of archaeological sites and proximity of these sites to the pollen samples locations. From the general framework structuring the vegetation cover changes for both study areas stand out the major steps of the development of natural landscapes during the Holocene. Nevertheless, regional distinctive features were identified; both in terms of chronology of the vegetation cover changes, than in terms of relative proportions of taxa in the vegetation estimates. Some singularity must be recognized in the Holocene landscape evolution for each of these two parts of north-western France. We noticed distinct rates and rhythms, concerning the anthropogenic impact on the Paris Basin and the Armorican Massif estimated vegetation covers, from the late Mesolithic to the late Neolithic. On the other hand, we found similarities between both areas concerning the estimated proportions for anthropogenic markers taxa between the late Neolithic and the late Bronze Age. Clear links between changes observed in human impact on the vegetation cover estimates and those recorded in the density of archaeological sites were also noticed.

Saltwater Forest is a Dacrydium cupressinum-dominated lowland forest covering 9000 ha in south Westland, South Island, New Zealand. Four thousand hectares is managed for sustainable production of indigenous timber. The aim of this study was to provide an integrated analysis of soils, soil-landform relationships, and soil-vegetation relationships at broad and detailed scales. The broad scale understandings provide a framework in which existing or future studies can be placed and the detailed studies elucidate sources of soil and forest variability. Glacial landforms dominate. They include late Pleistocene lateral, terminal and ablation moraines, and outwash aggradation and degradation terraces. Deposits and landforms from six glacial advances have been recognised ranging from latest Last (Otira) Glaciation to Penultimate (Waimea) Glaciation. The absolute ages of landforms were established by analysis of the thickness and soil stratigraphy of loess coverbeds, augmented with radiocarbon dating and phytolith and pollen analysis. In the prevailing high rainfall of Westland soil formation is rapid. The rate of loess accretion in Saltwater Forest (ca. 30 mm ka⁻¹) has been low enough that soil formation and loess accretion took place contemporaneously. Soils formed in this manner are known as upbuilding soils. The significant difference between upbuilding pedogenesis and pedogenesis in a topdown sense into an existing sediment body is that each subsoil increment of an upbuilding soil has experienced processes of all horizons above. In Saltwater Forest subsoils of upbuilding soils are strongly altered because they have experienced the extremely acid environment of the soil surface at some earlier time. Some soil chronosequence studies in Westland have included upbuilding soils formed in loess as the older members of the sequence. Rates and types of processes inferred from these soils should be reviewed because upbuilding is a different pedogenic pathway to topdown pedogenesis. Landform age and morphology were used as a primary stratification for a study of the soil pattern and nature of soil variability in the 4000 ha production area of Saltwater Forest. The age of landforms (> 14 ka) and rapid soil formation mean that soils are uniformly strongly weathered and leached. Soils include Humic Organic Soils, Perch-gley Podzols, Acid Gley Soils, Allophanic Brown Soils, and Orthic or Pan Podzols. The major influence on the nature of soils is site hydrology which is determined by macroscale features of landforms (slope, relief, drainage density), mesoscale effects related to position on landforms, and microscale influences determined by microtopography and individual tree effects. Much of the soil variability arises at microscales so that it is not possible to map areas of uniform soils at practical map scales. The distribution of soil variability across spatial scales, in relation to the intensity of forest management, dictates that it is most appropriate to map soil complexes with boundaries coinciding with landforms. Disturbance of canopy trees is an important agent in forest dynamics. The frequency of forest disturbance in the production area of Saltwater Forest varies in a systematic way among landforms in accord with changes in abundance of different soils. The frequency of forest turnover is highest on landforms with the greatest abundance of extremely poorly-drained Organic Soils. As the abundance of better-drained soils increases the frequency of forest turnover declines. Changes in turnover frequency are reflected in the mean size and density of canopy trees (Dacrydium cupressinum) among landforms. Terrace and ablation moraine landforms with the greatest abundance of extremely poorly-drained soils have on average the smallest trees growing most densely. The steep lateral moraines, characterised by well drained soils, have fewer, larger trees. The changes manifested at the landform scale are an integration of processes operating over much shorter range as a result of short-range soil variability. The systematic changes in forest structure and turnover frequency among landforms and soils have important implications for sustainable forest management.

Saltwater Forest is a Dacrydium cupressinum-dominated lowland forest covering 9000 ha in south Westland, South Island, New Zealand. Four thousand hectares is managed for sustainable production of indigenous timber. The aim of this study was to provide an integrated analysis of soils, soil-landform relationships, and soil-vegetation relationships at broad and detailed scales. The broad scale understandings provide a framework in which existing or future studies can be placed and the detailed studies elucidate sources of soil and forest variability. Glacial landforms dominate. They include late Pleistocene lateral, terminal and ablation moraines, and outwash aggradation and degradation terraces. Deposits and landforms from six glacial advances have been recognised ranging from latest Last (Otira) Glaciation to Penultimate (Waimea) Glaciation. The absolute ages of landforms were established by analysis of the thickness and soil stratigraphy of loess coverbeds, augmented with radiocarbon dating and phytolith and pollen analysis. In the prevailing high rainfall of Westland soil formation is rapid. The rate of loess accretion in Saltwater Forest (ca. 30 mm ka⁻¹) has been low enough that soil formation and loess accretion took place contemporaneously. Soils formed in this manner are known as upbuilding soils. The significant difference between upbuilding pedogenesis and pedogenesis in a topdown sense into an existing sediment body is that each subsoil increment of an upbuilding soil has experienced processes of all horizons above. In Saltwater Forest subsoils of upbuilding soils are strongly altered because they have experienced the extremely acid environment of the soil surface at some earlier time. Some soil chronosequence studies in Westland have included upbuilding soils formed in loess as the older members of the sequence. Rates and types of processes inferred from these soils should be reviewed because upbuilding is a different pedogenic pathway to topdown pedogenesis. Landform age and morphology were used as a primary stratification for a study of the soil pattern and nature of soil variability in the 4000 ha production area of Saltwater Forest. The age of landforms (> 14 ka) and rapid soil formation mean that soils are uniformly strongly weathered and leached. Soils include Humic Organic Soils, Perch-gley Podzols, Acid Gley Soils, Allophanic Brown Soils, and Orthic or Pan Podzols. The major influence on the nature of soils is site hydrology which is determined by macroscale features of landforms (slope, relief, drainage density), mesoscale effects related to position on landforms, and microscale influences determined by microtopography and individual tree effects. Much of the soil variability arises at microscales so that it is not possible to map areas of uniform soils at practical map scales. The distribution of soil variability across spatial scales, in relation to the intensity of forest management, dictates that it is most appropriate to map soil complexes with boundaries coinciding with landforms. Disturbance of canopy trees is an important agent in forest dynamics. The frequency of forest disturbance in the production area of Saltwater Forest varies in a systematic way among landforms in accord with changes in abundance of different soils. The frequency of forest turnover is highest on landforms with the greatest abundance of extremely poorly-drained Organic Soils. As the abundance of better-drained soils increases the frequency of forest turnover declines. Changes in turnover frequency are reflected in the mean size and density of canopy trees (Dacrydium cupressinum) among landforms. Terrace and ablation moraine landforms with the greatest abundance of extremely poorly-drained soils have on average the smallest trees growing most densely. The steep lateral moraines, characterised by well drained soils, have fewer, larger trees. The changes manifested at the landform scale are an integration of processes operating over much shorter range as a result of short-range soil variability. The systematic changes in forest structure and turnover frequency among landforms and soils have important implications for sustainable forest management.